## No bets were placed for February 14
## No results available
## No bets yesterday
Remember: Points = UNDERS | Rebounds = OVERS | Assists = OVERS
## Picks will be available after 10 AM ET line scrape
## Picks will be available after 10 AM ET line scrape
Update (2026-01-31): After fixing lookahead bias in the backtest, the matchup filter shows no meaningful edge. Earlier results (+6-12% ROI) were inflated because we used current team rankings on historical data. With proper dynamic rankings calculated as of each game date, results are essentially break-even.
Bottom Line: After correcting for lookahead bias, the matchup filter shows no reliable edge for rebounds or assists. Rebounds are marginally better vs hard matchups (56.2% vs 54.7% win rate) but still near break-even. Stick to points UNDERS only.
Investigation (2026-02-15): Can we find edge on alternate player point lines? Books offer multiple lines per player (e.g., Under 19.5, Under 24.5, Under 29.5) at different odds. We tested whether our model could exploit these.
Bottom Line: Alternate lines are priced efficiently. Win rates increase with higher edges, but the odds (juice) increase proportionally. Even at 8+ edge (86% WR), ROI is ~0%. The books know exactly how likely an Under is at each alternate line. No exploitable edge found.
THE RULE IS SIMPLE:
| Prop Type | Recommendation | Why |
|---|---|---|
| POINTS | UNDERS only | Model underpredicts scoring (+7-11% ROI) |
| REBOUNDS | NOT RECOMMENDED | No reliable edge after proper backtest |
| ASSISTS | NOT RECOMMENDED | No reliable edge after proper backtest |
| Prop | Bet Type | Win Rate | ROI | Strategy |
|---|---|---|---|---|
| Points | Under | 55-57% | +10-15% | All matchups, 2+ edge |
| Rebounds | Under | 63% | +12.6% | Hard matchups only |
| Assists | Under | 61% | +7.4% | Hard matchups only |
Hard Matchups = Top 8 defensive teams (rolling 45-day calculation)
Elastic Net Regression - A regularized linear model that predicts player stats using recent performance data.
| Prop Type | Bet Direction | Status | Win Rate | ROI | Filter |
|---|---|---|---|---|---|
| Points | UNDERS | Production | 55-57% | +10-15% | All matchups |
| Rebounds | UNDERS | Production | 63% | +12.6% | Hard matchups only |
| Assists | UNDERS | Production | 61% | +7.4% | Hard matchups only |
Hard matchups = Top 8 defensive teams (recalculated daily from last 45 days)
The points model systematically underpredicts player scoring, which makes UNDERS profitable.
Features used (12 total):
| Feature | Description |
|---|---|
| roll_pts_3/5/7/9 | Rolling points average over 3, 5, 7, 9 games |
| roll_mp_3/5 | Rolling minutes played |
| roll_usg_3/5 | Rolling usage percentage |
| roll_3pa_3 | Rolling 3-point attempts |
| career_avg_pts | Career points average |
| roll_pts_sd_5 | Scoring variability (5-game std dev) |
| days_rest | Days since last game |
Backtest results (walk-forward):
| Strategy | Win Rate | ROI |
|---|---|---|
| All bets | 48% | -8% |
| Overs only | 44% | -15% |
| Unders only | 55% | +18% |
| Unders + 4pt edge | 57% | +25% |
Betting rule: UNDERS only with 3+ point edge
Key Discovery (2026-01-31): Sportsbooks don’t fully adjust lines for opponent defensive quality. UNDERS are profitable when facing hard matchups (top-8 rebounding defenses).
Features used (12 total):
| Feature | Description |
|---|---|
| roll_trb_3/5/7/9 | Rolling total rebounds over 3, 5, 7, 9 games |
| roll_mp_3/5 | Rolling minutes played |
| roll_orb_5, roll_drb_5 | Offensive/defensive rebounds |
| career_avg_trb | Career rebounds average |
| roll_trb_sd_5 | Rebounding variability |
| reb_per_min | Rebounds per minute rate |
| days_rest | Days since last game |
Backtest results with DYNAMIC matchup filter (no lookahead bias):
| Matchup Type | Bets | Win Rate | ROI |
|---|---|---|---|
| All matchups | 19,218 | 55.0% | -2.3% |
| Hard matchups | 3,953 | 56.2% | +0.03% |
| Other matchups | 15,265 | 54.7% | -2.9% |
Result: Marginally better win rate vs hard matchups, but essentially break-even. Not recommended.
Backtest results with DYNAMIC matchup filter (no lookahead bias):
| Matchup Type | Bets | Win Rate | ROI |
|---|---|---|---|
| All matchups | 12,866 | 52.8% | -6.1% |
| Hard matchups | 2,950 | 53.4% | -5.2% |
| Other matchups | 9,916 | 52.6% | -6.4% |
Result: No meaningful difference between hard and other matchups. Not recommended.
Features used (11 total):
| Feature | Description |
|---|---|
| roll_3pm_3/5/7/9 | Rolling 3-pointers made over 3, 5, 7, 9 games |
| roll_mp_3/5 | Rolling minutes played |
| roll_3pa_5 | Rolling 3-point attempts (volume) |
| roll_3p_pct_5 | Rolling 3-point shooting percentage |
| career_avg_3pm | Career 3-pointers made average |
| roll_3pm_sd_5 | 3-point variability (5-game std dev) |
| days_rest | Days since last game |
Status: New prop type - backtest pending historical data collection.
| Rule | Requirement |
|---|---|
| Bet Type | UNDERS only |
| Min Edge | 3+ points |
| Max Odds | -150 or better |
| HIGH conf | 4+ point edge |
NOT RECOMMENDED - Dynamic backtest shows no reliable edge (+0.03% ROI)
NOT RECOMMENDED - Dynamic backtest shows no edge (-5.2% ROI)
NEW - Backtest pending. UNDERS only with 1+ point edge (preliminary).
Generated 2026-02-15 13:01 ET